Shuichiro Ogawa
日本語

Notes · updated 2026-08-18

Novice Group Work with a Shared AI Agent: A Literature Map of CSCL and AI-Supported Collaborative Learning (2026)

An integrative summary of 48 publications (40 from the initial collection plus 8 added by a same-day verification sweep; peer-reviewed papers, reviews, and major preprints) collected through a lightweight scoping review on the design and best practices of introducing a single shared AI agent (one conversational agent, chatbot, pedagogical agent, robot, or LLM-based agent per group) into group work with novice learners. Full bibliographic details appear in the “References” section below (with DOIs/URLs). The internal corpus with the provenance ledger is at source/review/shared-ai-agent-groupwork/papers.md (repository-internal, not published). For the broader research landscape of AI and education, see ai-education-learning-sciences-trends; for the literature map on cognitive offloading, see ai-cognitive-offloading-learning; for design education curricula, see design-education-ai-adaptation. Collection pipeline: source-researcher (profile: scholarly) followed by paper-screening-agent. Protocol: .claude/rules/collection-protocol.md (zero fabrication, provenance tracking). On 2026-08-18 a verification sweep (direct ERIC/DOAJ search, replication search for interaction drift, preprint publication-status checks) ran as three parallel agents, added 8 items, and updated the verdicts in the “Gaps” section.

Survey Metadata

  • Collection date: 2026-08-18 (initial collection and verification sweep on the same day) / Count: 48 (initial: 41 candidates, 1 excluded, 40 kept; sweep: 8 added; lightweight scoping prioritizes minimizing omissions)
  • Venue distribution: Springer venues (ijCSCL, IJAIED, Educational Psychology Review, ETR&D), Wiley (BJET, JCAL), Computers & Education family, arXiv, Frontiers, IJCAI/ICCE/AIED
  • Confidence notes: 7 non-peer-reviewed preprints (sag34, sag35, sag37, sag41, sag44, sag45, sag46) are flagged non-peer. sag26 and sag33 were upgraded after confirming their peer-reviewed versions (AIED 2024 / AIED 2026, LNCS). Zero retractions or predatory venues

TL;DR

Research on introducing AI into groups did not start with LLMs: it has roughly twenty years of accumulated work under ACLS (adaptive collaborative learning support, which models a group’s dialogue state and intervenes adaptively), including established findings on “when, to whom, and what to say” (triggering productive talk, directed vs. undirected interventions). LLM-era empirical work (2022 onward) inherits this lineage and adds two new findings. First, embedding pedagogical strategy into LLMs makes adaptive scaffolding easier to implement. Second, a shared chatbot can absorb student-student interaction: network analysis shows students interact more with the AI than with each other. The verification sweep, however, suggests this absorption is condition-dependent: a real-classroom study found that a single shared agent promoted integrated group reasoning, while fragmentation occurred in the one-AI-per-student condition. Best practice is converging on “design the AI as a facilitator rather than an answer source, let it intervene selectively based on detected dialogue state, and monitor both student-student interaction and AI reliance,” but novice-specific intervention design principles and, within education, interventions that directly target free-riding remain thin.

Key Findings by Cluster

A. Foundations: collaboration does not happen by itself (8 papers)

  • The starting points are Dillenbourg’s definition of collaborative learning (sag01) and the warning by Kreijns et al. (sag02): the assumption that “providing an environment makes interaction happen naturally” is the classic pitfall of CSCL, and it applies directly to today’s expectation that “placing an AI will improve collaboration.”
  • The structuring device is the collaboration script (specifying roles, sequences, and activities), with conceptual analyses (sag03, sag04) and a meta-analysis (sag05) quantifying its effects on collaboration skills and domain learning.
  • Group awareness research (sag06, sag07, sag08) showed that visualization tools alone are insufficient; what matters is how they connect to intervention. This is the direct precursor of AI agents that detect group state and intervene.

B. ACLS: the pre-LLM lineage of adaptive intervention (9 papers)

  • Walker et al.’s CTRL framework (sag09) and the outlook by Rummel et al. (sag10) formalized ACLS as a research program; the shift from fixed scripts toward adaptive intervention was established here.
  • The conversational-agent lineage runs from Kumar et al.’s pair-learning support (sag11) through the Bazaar architecture (sag12) to the Tegos MentorChat line (sag13, sag14, sag15). Agent interventions increase explanatory talk and predict knowledge gains (sag14), and effects differ depending on whether interventions are directed at a specific learner or undirected to the whole group (sag15).
  • Field comparisons in university classrooms (sag18) and the MOOC extension PeerTalk (sag29) demonstrate operational feasibility in real classes.

C. Role typology and embodiment: tutor, facilitator, or peer (8 papers)

  • The systematic review (sag40; 46 empirical studies, 2014–2025) organizes agent roles into cognitive scaffolding, social facilitation, and instructional orchestration. It is the most recent map of role types.
  • A design experiment in a vocational school (sag48) built a chatbot as a shared tutor and teammate for triads. The authors themselves name the short four-lesson trial as a limitation, corroborating the lack of long-term field studies (see Gaps) in the practitioners’ own words.
  • Single vs. multiple agents: Hayashi’s experiments (sag16, sag17) tested splitting roles across agents and found that how suggestions are divided affects comprehension. A position paper (sag41) sketches multi-persona futures.
  • Embodiment options include social robots (sag27, small-group facilitation in higher education) and virtual robots (sag28, K-12 online discussion). Surface design such as voice accent shapes trust and role perception (sag35), showing anthropomorphization is not neutral.

D. SSRL x AI: supporting the group’s own regulation (5 papers)

  • Järvelä et al.’s HASRL model (sag30) frames how AI can support SSRL (socially shared regulation of learning: a group jointly planning, monitoring, and regulating its learning); Edwards et al. (sag32) concretize the agent design.
  • Learners perceive AI as an active agent supporting SSRL (sag31). Meanwhile, an experiment (sag33, published at AIED 2026) reports that with GenAI available, group regulation itself shifts from socially shared to hybrid human-AI co-regulation, suggesting a shared agent reshapes the group’s regulation structure.
  • Implementations have reached high schools: the SSRL-support chatbot “RegTogether” (sag49) assisted high-school groups’ joint regulation, with the experimental group outperforming controls on group task performance (the sharing configuration and study duration remain [requires primary verification]).

E. LLM-era empirical studies of shared agents (9 papers)

  • CollaBot (sag21), combining LLM + RAG + a contingent-teaching model, provides adaptive scaffolding on cognitive, metacognitive, and social layers, with reported effects on learning outcomes and self-efficacy.
  • Design principles are emerging: a design-based research study derived 4 design components and 10 principles aimed at collective efficacy (sag22).
  • Failure findings are instructive: a prompt-only LLM struggles with multi-party dialogue processing and the injection of pedagogical strategy (sag24). In response, a pedagogy-guided framework detects discussion phases and decides intervention timing (sag25, grounded in Community of Inquiry theory).
  • Examples of “supporting without giving answers” include situated reflection triggers generated by GenAI (sag26) and theory-guided scaffolding designed to avoid over-reliance (Copa, sag34, 33 high-school pairs). A simulation-based method pre-tests scaffolds with LLM multi-agent systems (sag37).
  • Group composition has become a variable too. A quasi-experiment (sag47, 196 undergraduates) composed homogeneous or heterogeneous groups by prior-knowledge level and compared collaborative problem solving with ChatGPT, finding an interaction effect: heterogeneous grouping helped low-prior-knowledge learners. It is the first study in this corpus to treat prior knowledge as an explicit variable of shared-AI group work.

F. Risks and the adopter’s perspective (8 papers)

  • Network analysis (sag23) showed that in groups with a shared GAI chatbot, student-chatbot interaction significantly exceeds student-student interaction: direct evidence that a shared AI can become the hub of collaboration and thin out peer dialogue.
  • Follow-up studies from the verification sweep show this absorption is condition-dependent. In a real-classroom study with 8th graders (sag44), the single shared-agent condition promoted integrated group reasoning, and fragmentation occurred instead in the one-AI-per-student condition. A role-comparison experiment (sag43) found peer-type AI disrupted group flow while facilitator-type AI had limited impact, and a randomized controlled experiment (sag42) showed that fast, uniform AI responses can reduce student-student discussion. Absorption is driven not by “the AI being shared” but by the role (peer vs. facilitator), response speed and uniformity, and shared vs. per-student configuration. A strict replication of sag23 (the same network-analysis comparison of interaction frequencies) still does not exist.
  • Measurement studies of participation balance have appeared: a persona-manipulation experiment (sag45, 92 triads) reported that participation equality was maintained under both supportive and antagonistic AI teammates, and an “inclusion analytics” framework (sag46) measures participation balance and emotional climate. Both, however, measure balance rather than intervene to create it.
  • A scoping review (sag38) organizes the risks of GenAI in group work: over-reliance, free-riding, and difficulty in fair assessment of contributions.
  • Teacher interviews (sag36) show expectations for engagement alongside concerns about autonomy, trust, anthropomorphization, and pedagogical alignment. Deployment design must include teacher orchestration.

Design Implications for Novice Groups (the current state of best practice)

Mapping the ledger’s findings onto deployment decisions:

  1. Placing an AI does not improve collaboration by itself. Interaction must be structured (sag02). Design roles and sequences first (collaboration scripts, sag03, sag05) and layer AI intervention on top.
  2. Selective intervention based on detected dialogue state, not constant intervention. Supported jointly by the ACLS framework (sag09, sag10) and LLM-era intervention-timing work (sag25).
  3. Design the addressee of interventions. Effects differ between whole-group and individual-directed interventions (sag15); directed prompts to quiet novices can serve participation balance.
  4. Make the agent a trigger of productive talk, not an answer source. Interventions eliciting explanatory reasoning (sag13, sag14), reflection triggers (sag26), and scaffolding designed against over-reliance (sag34) are the implemented examples.
  5. Monitor AI-centered interaction drift. A shared chatbot can absorb peer interaction (sag23), but the drift is condition-dependent: a counter-example shows the single-shared configuration promoted integrated reasoning while fragmentation arose with per-student AI (sag44); facilitator-type roles are safer than peer-type (sag43), and instant, uniform responses tend to reduce discussion (sag42). Use group awareness findings (sag06, sag07) to visualize and preserve the ratio of student-student dialogue.
  6. Make SSRL support an explicit design goal. Support joint planning, monitoring, and reflection (sag30, sag32), while assuming AI availability hybridizes the regulation structure itself (sag33).
  7. A bare LLM prompt is not enough. Multi-party dialogue handling and pedagogical strategy must be engineered (sag24), via embedded teaching models (sag21) or pedagogy-guided intervention control (sag25).
  8. Include teacher orchestration in the design. Configurability (sag29), classroom intervention-strategy comparisons (sag18), and teachers’ concerns (sag36) support positioning the agent as assisting, not replacing, the teacher.
  9. Build risk management into assessment design. Over-reliance, free-riding, and contribution-assessment difficulties (sag38) call for assessment that keeps individual contributions visible.

Gaps (unmet questions; verdicts updated by the verification sweep)

  • Novice-specific design principles: the sweep found an experiment treating prior-knowledge level as an explicit variable of shared-AI group work (sag47), so the initial claim that “no such study exists” is narrowed and corrected. That study analyzes group composition, however; no study yet offers novice-specific intervention design principles (when, to whom, and how to intervene).
  • Interventions targeting participation balance and free-riding: within education, controlled experiments now measure balance (sag45) and a measurement framework exists (sag46), but no controlled study yet makes balance or free-riding mitigation the direct goal of the intervention. In the adjacent domain of adult group deliberation, pre-registered RCTs already target participation equity (Alsobay et al. 2025 improved information sharing; Parisi et al. 2026 reported no improvement in participation equity; both outside the corpus scope). The gap claim is therefore restricted to educational settings.
  • Maturity of LLM-era evidence: of the six items that were preprints at initial collection, two (sag26, sag33) were confirmed as peer-reviewed (AIED 2024 / AIED 2026) and upgraded; no retractions or substantial critiques were found for any of the six. Seven items remain non-peer-reviewed (including sweep additions), and most experiments are short and small. The lack of semester-scale field studies stands, and the authors of a design experiment name the short trial as a limitation themselves (sag48).
  • Connecting ACLS classics to LLM implementations: findings such as directed/undirected interventions (sag15) and APT interventions (sag14) have not been systematically ported to LLM-based designs.
  • Coverage limitation: the sweep searched ERIC directly via its API and checked all 28 citations of sag23 on Semantic Scholar (resolving the initial limitation). DOAJ remained unreachable (HTTP 403), so additional OA-journal candidates are unconfirmed.

Unverified Items

  • sag48 (Burkhard et al. 2022, CELDA): final confirmation that no DOI exists for the proceedings is pending [requires primary verification].
  • sag49 (Kim, Jeong, Lim 2026): the AI sharing configuration (one agent per group vs. individual access) and study duration [requires primary verification].
  • sag43 (Mukhopadhyay et al. 2026, CHI): whether participants were students or general adults [requires primary verification].
  • sag46 (Choi & Nixon 2026): experimental details such as control groups and sample size [requires primary verification].
  • Outside the corpus: Kim, J. et al. 2026, “Student-AI interaction patterns in collaborative problem solving using ordered network analysis” (Interactive Learning Environments, https://doi.org/10.1080/10494820.2026.2664072) likely bears directly on the interaction-drift replication question, but the full text was unreachable and its content is unverified [requires primary verification].
  • Resolved: the LNCS volume number of sag17 was confirmed as 8474 during the sweep (via DBLP).

References

All bibliographic records were reached and verified on 2026-08-18 via Crossref, OpenAlex, arXiv, or publisher pages.

  1. Dillenbourg, P. 1999. What Do You Mean by Collaborative Learning? In Collaborative-learning: Cognitive and Computational Approaches (pp.1–19). Elsevier. https://tecfa.unige.ch/tecfa/teaching/aei/papiers/Dillenbourg.pdf
  2. Kreijns, K., Kirschner, P. A., & Jochems, W. 2003. Identifying the pitfalls for social interaction in computer-supported collaborative learning environments. Computers in Human Behavior, 19(3), 335–353. https://doi.org/10.1016/s0747-5632(02)00057-2
  3. Kollar, I., Fischer, F., & Hesse, F. W. 2006. Collaboration Scripts – A Conceptual Analysis. Educational Psychology Review, 18(2), 159–185. https://doi.org/10.1007/s10648-006-9007-2
  4. Kobbe, L., Weinberger, A., Dillenbourg, P., et al. 2007. Specifying computer-supported collaboration scripts. International Journal of Computer-Supported Collaborative Learning, 2(2–3). https://doi.org/10.1007/s11412-007-9014-4
  5. Vogel, F., Wecker, C., Kollar, I., & Fischer, F. 2017. Socio-Cognitive Scaffolding with Computer-Supported Collaboration Scripts: a Meta-Analysis. Educational Psychology Review, 29(3), 477–511. https://doi.org/10.1007/s10648-016-9361-7
  6. Janssen, J., & Bodemer, D. 2013. Coordinated Computer-Supported Collaborative Learning: Awareness and Awareness Tools. Educational Psychologist, 48(1), 40–55. https://doi.org/10.1080/00461520.2012.749153
  7. Janssen, J., Erkens, G., & Kirschner, P. A. 2011. Group awareness tools: It’s what you do with it that matters. Computers in Human Behavior, 27(3), 1046–1058. https://doi.org/10.1016/j.chb.2010.06.002
  8. Chen, D., Zhang, Y., & Lin, Y. 2022. Group Awareness and Group Awareness Tools in Computer-Supported Collaborative Learning: A Literature Review. ISET 2022, 18–22. https://doi.org/10.1109/iset55194.2022.00013
  9. Walker, E., Rummel, N., & Koedinger, K. R. 2009. CTRL: A research framework for providing adaptive collaborative learning support. User Modeling and User-Adapted Interaction, 19(5), 387–431. https://doi.org/10.1007/s11257-009-9069-1
  10. Rummel, N., Walker, E., & Aleven, V. 2016. Different Futures of Adaptive Collaborative Learning Support. International Journal of Artificial Intelligence in Education, 26(2), 784–795. https://doi.org/10.1007/s40593-016-0102-3
  11. Kumar, R., Rosé, C. P., Wang, Y.-C., Joshi, M., & Robinson, A. L. 2007. Tutorial Dialogue as Adaptive Collaborative Learning Support. AIED 2007 (IOS Press), 383–390. http://www.cs.cmu.edu/~maheshj/pubs/kumar+rose+wang+joshi+robinson.aied2007.pdf
  12. Adamson, D., Dyke, G., Jang, H., & Rosé, C. P. 2014. Towards an Agile Approach to Adapting Dynamic Collaboration Support to Student Needs. International Journal of Artificial Intelligence in Education, 24(1), 92–124. https://doi.org/10.1007/s40593-013-0012-6
  13. Tegos, S., Demetriadis, S., & Tsiatsos, T. 2014. A Configurable Conversational Agent to Trigger Students’ Productive Dialogue: A Pilot Study in the CALL Domain. International Journal of Artificial Intelligence in Education, 24(1), 62–91. https://doi.org/10.1007/s40593-013-0007-3
  14. Tegos, S., Demetriadis, S., & Karakostas, A. 2015. Promoting academically productive talk with conversational agent interventions in collaborative learning settings. Computers & Education, 87, 309–325. https://doi.org/10.1016/j.compedu.2015.07.014
  15. Tegos, S., Demetriadis, S., Papadopoulos, P. M., & Weinberger, A. 2016. Conversational agents for academically productive talk: a comparison of directed and undirected agent interventions. International Journal of Computer-Supported Collaborative Learning, 11(4), 417–440. https://doi.org/10.1007/s11412-016-9246-2
  16. Hayashi, Y. 2019. Multiple pedagogical conversational agents to support learner-learner collaborative learning: Effects of splitting suggestion types. Cognitive Systems Research, 54, 246–257. https://doi.org/10.1016/j.cogsys.2018.04.005
  17. Hayashi, Y. 2014. Togetherness: Multiple Pedagogical Conversational Agents as Companions in Collaborative Learning. ITS 2014, LNCS 8474, 114–123. https://doi.org/10.1007/978-3-319-07221-0_14
  18. Michos, K., Asensio-Pérez, J. I., Dimitriadis, Y., et al. 2020. Design of Conversational Agents for CSCL: Comparing Two Types of Agent Intervention Strategies in a University Classroom. EC-TEL 2020, LNCS. https://doi.org/10.1007/978-3-030-57717-9_16
  19. Araujo, A. de, Papadopoulos, P. M., McKenney, S., & de Jong, T. 2024. A learning analytics-based collaborative conversational agent to foster productive dialogue in inquiry learning. Journal of Computer Assisted Learning, 40(6). https://doi.org/10.1111/jcal.13007
  20. Araujo, A. de, Papadopoulos, P. M., McKenney, S., & de Jong, T. 2025. Investigating the Impact of a Collaborative Conversational Agent on Dialogue Productivity and Knowledge Acquisition. International Journal of Artificial Intelligence in Education, 35(4), 2254–2280. https://doi.org/10.1007/s40593-025-00469-7
  21. Hu, W., Gong, R., Wu, S., & Li, Y. 2025. A conversational agent based on contingent teaching model to support collaborative learning activities. Educational Technology Research and Development, 73(5). https://doi.org/10.1007/s11423-025-10526-6
  22. Kim, M., & Lim, C. 2025. AI chatbot design principles to enhance the collective efficacy in collaborative learning. Asia Pacific Education Review, 26(4). https://doi.org/10.1007/s12564-025-10080-z
  23. Feng, S. 2025. Group interaction patterns in generative AI-supported collaborative problem solving: Network analysis of the interactions among students and a GAI chatbot. British Journal of Educational Technology, 56(5), 2125–2145. https://doi.org/10.1111/bjet.13611
  24. An, S., Li, Y., Yang, Y., Ma, Y., Cheng, G., & Xu, G. 2024. Developing an LLM-Empowered Agent to Enhance Student Collaborative Learning Through Group Discussion. ICCE 2024. https://doi.org/10.58459/icce.2024.4953
  25. Yang, Q., Yang, Y., An, S., Hao, T., & Xu, G. 2025. LLM-based Collaborative Agents with Pedagogy-guided Interaction Modeling for Timely Instructive Feedback Generation in Task-oriented Group Discussions. IJCAI 2025, 9972–9980. https://doi.org/10.24963/ijcai.2025/1108
  26. Naik, A., Yin, J. R., Kamath, A., et al. 2024. Generating Situated Reflection Triggers about Alternative Solution Paths: A Case Study of Generative AI for Computer-Supported Collaborative Learning. AIED 2024, LNCS 14829, 46–59. https://doi.org/10.1007/978-3-031-64302-6_4
  27. Rosenberg-Kima, R. B., Koren, Y., & Gordon, G. 2020. Robot-Supported Collaborative Learning (RSCL): Social Robots as Teaching Assistants for Higher Education Small Group Facilitation. Frontiers in Robotics and AI, 6, 148. https://doi.org/10.3389/frobt.2019.00148
  28. Mizrahi, E., Danzig, N., & Gordon, G. 2022. vRobotator: A Virtual Robot Facilitator of Small Group Discussions for K-12. Proceedings of the ACM on Human-Computer Interaction, 6(CSCW2), 1–22. https://doi.org/10.1145/3555573
  29. Tegos, S., Mavridis, A., & Demetriadis, S. 2021. Agent-Supported Peer Collaboration in MOOCs. Frontiers in Artificial Intelligence, 4, 710856. https://doi.org/10.3389/frai.2021.710856
  30. Järvelä, S., Nguyen, A., & Hadwin, A. 2023. Human and artificial intelligence collaboration for socially shared regulation in learning. British Journal of Educational Technology, 54(5), 1057–1076. https://doi.org/10.1111/bjet.13325
  31. Kim, J., Detrick, R., Yu, S., Song, Y., Bol, L., & Li, N. 2025. Socially shared regulation of learning and artificial intelligence: Opportunities to support socially shared regulation. Education and Information Technologies, 30(9), 11483–11521. https://doi.org/10.1007/s10639-024-13187-9
  32. Edwards, J., Nguyen, A., Lämsä, J., et al. 2025. Human–AI collaboration: Designing artificial agents to facilitate socially shared regulation among learners. British Journal of Educational Technology, 56(2), 712–733. https://doi.org/10.1111/bjet.13534
  33. Zhang, Y., Meng, X., Feng, S., & Lin, J. 2026. Human-AI Collaboration Reconfigures Group Regulation from Socially Shared to Hybrid Co-Regulation. AIED 2026, LNCS 16585, 318–326. https://doi.org/10.1007/978-3-032-29770-9_35
  34. Cohn, C., Rayala, S., Guo, S., et al. 2026. A Theory-Guided LLM Pedagogical Agent for STEM+C Scaffolding Without Over-Reliance. arXiv:2605.30539 (preprint). https://arxiv.org/abs/2605.30539
  35. Ravi, P., Stevens, C., Hurt, B., et al. 2026. Exploring How Agent Voice Accents Shape Human-AI Collaboration in K-12 Group Learning. arXiv:2606.12805 (preprint). https://arxiv.org/abs/2606.12805
  36. Ravi, P., Stevens, C., Azevedo, B. F., et al. 2026. Exploring Teachers’ Perspectives on Using Conversational AI Agents for Group Collaboration. AIED 2026, LNCS. https://doi.org/10.1007/978-3-032-29763-1_43
  37. Wu, H., Zhang, L., & Lu, C. 2026. A Simulation-Based Method for Testing Collaborative Learning Scaffolds Using LLM-Based Multi-Agent Systems. arXiv:2604.11161 (preprint). https://arxiv.org/abs/2604.11161
  38. Wei, Y., & Perkins, M. 2026. Generative AI and student collaboration: a scoping review of group work processes, outcomes, and risks. International Journal for Educational Integrity, 22(1). https://doi.org/10.1007/s40979-026-00217-x
  39. Ba, S., Shi, X., Wu, S., & Lu, G. 2026. Artificial Intelligence Agents in Computer-Supported Collaborative Learning: A Systematic Literature Review. Computers and Education: Artificial Intelligence, 10, 100579. https://doi.org/10.1016/j.caeai.2026.100579
  40. Cox, S. R. 2023. The Use of Multiple Conversational Agent Interlocutors in Learning. Inter.HAI’23 workshop (preprint). https://arxiv.org/abs/2312.16534
  41. Han, R. 2026. Comparing AI chatbots and web-based synchronous collaborative writing tools, leveraging group cohesion, motivation, and performance. Acta Psychologica, 268, 107355. https://doi.org/10.1016/j.actpsy.2026.107355
  42. Mukhopadhyay, A., Salubre, K., Javed, H., Mehrotra, S., & Akash, K. 2026. Exploring The Impact of Proactive Generative AI Agent Roles In Time-Sensitive Collaborative Problem-Solving Tasks. CHI ‘26. https://doi.org/10.1145/3772318.3791592
  43. Zha, S., Liu, W., Qin, F., et al. 2026. How GenAI Mentor Configurations Shape Early Collaborative Dynamics. arXiv:2603.12600 (preprint). https://arxiv.org/abs/2603.12600
  44. Yan, L., Jin, Y., Zhao, L., Martinez-Maldonado, R., et al. 2025. Agentic AI as Undercover Teammates: Argumentative Knowledge Construction in Hybrid Human-AI Collaborative Learning. arXiv:2512.08933 (preprint). https://arxiv.org/abs/2512.08933
  45. Choi, J., & Nixon, N. 2026. Measuring Inclusion in Interaction: Inclusion Analytics for Human-AI Collaborative Learning. arXiv:2602.09269 (preprint). https://arxiv.org/abs/2602.09269
  46. Shulgina, G., Getman, A., Gulenkov, I., & Costley, J. 2025. Collaboration with AI: How group composition affects solution quality in problem solving. Innovations in Education and Teaching International. https://doi.org/10.1080/14703297.2025.2539779
  47. Burkhard, M., Seufert, S., Cetto, M., & Handschuh, S. 2022. Educational Chatbots for Collaborative Learning: Results of a Design Experiment in a Middle School. CELDA 2022. https://eric.ed.gov/?id=ED626892
  48. Kim, M., Jeong, S., & Lim, C. 2026. Leveraging an AI Chatbot to Support Socially Shared Regulation in Collaborative Project-Based Learning. Asia Pacific Education Review. https://doi.org/10.1007/s12564-026-10114-0

← All Notes · Home